The present invention relates to directional interpolation of an image, and more particularly, to a method for performing directional interpolation using frequency information (e.g., Discrete Cosine Transform (DCT) information) and a related device.
Traditional interpolation methods assume that a two-dimensional image to be interpolated consists of a plurality of one-dimensional images, and therefore, the two-dimensional image is interpolated vertically or horizontally. Such treatment may introduce unwanted zigzag edges of objects in the image. Some directional interpolation methods have been proposed to solve the above-mentioned problem.
As shown in FIG. 1, the ELA algorithm (ref. T. Doyle and M. Looymans, Progressive scan conversion using edge information. In: L. Chiariglione, ed. Signal Processing of HDTV II, Elsevier Science Publishers B. V., North-Holland, pp. 711-721, 1990) is typical of the directional interpolation methods known in the art. Firstly, a direction corresponding to a pair of pixels having a minimum difference between their pixel values is selected from among three directions 131, 132, 133. The pixel values typically represent brightness and/or color levels of each pixel. For example, pixels 111, 123 are substantially equivalent and the difference between the values of the pixels 111, 123 is therefore smaller than the difference between the values of pixels 112, 122 and smaller than the difference between the values of pixels 113, 121. Because of this, directional interpolation should be performed along the direction 131. That is, the interpolated pixel 152 is derived from averaging the values of the pixels 111, 123. As shown in FIG.2, pixels 201, 202, 203 are interpolated vertically while pixels 211, 212, 213 are interpolated along an edge 215 of an object 210 of an interpolated image. The shading shown for each pixel in FIG.2 denotes a hypothetical pixel value after interpolation. It is evident in this example that the zigzag edge problem is solved by using directional interpolation. For more information about variations of the ELA algorithm, please refer to U.S. Pat. Nos. 5,742,348 and 6,133,957.
However, the above-mentioned directional interpolation algorithm introduces a new problem of faulty treatment of images with noise. Specifically, noise will decrease the correctness of the selection of the edge direction using the pixel values. Therefore, if the interpolation direction is determined according to the incorrectly selected edge direction, the directional interpolation will not work properly.